Those cheap Bluetooth Tesla coils on AliExpress use A2DP audio as their control signal - a comparator converts the audio into spark pulses. I built a browser app that generates the right square wave audio to drive them. But you don't need a coil to play with it.
159 tracks, a step sequencer, tone generators, and a real-time lightning visualiser driven by FFT. Works on phone or desktop, no install.
I'm also building a pipeline to convert any song into something a coil can play - monophonic square waves, 20Hz-2kHz. Using Demucs for stem separation and Basic Pitch for note detection. The annoying problem: ML models keep interpreting a square wave's harmonics as separate notes, so there's a bunch of post-processing to filter those out.
All Web Audio API, all client-side. No server in the audio path.
Those cheap Bluetooth Tesla coils on AliExpress use A2DP audio as their control signal - a comparator converts the audio into spark pulses. I built a browser app that generates the right square wave audio to drive them. But you don't need a coil to play with it.
159 tracks, a step sequencer, tone generators, and a real-time lightning visualiser driven by FFT. Works on phone or desktop, no install.
I'm also building a pipeline to convert any song into something a coil can play - monophonic square waves, 20Hz-2kHz. Using Demucs for stem separation and Basic Pitch for note detection. The annoying problem: ML models keep interpreting a square wave's harmonics as separate notes, so there's a bunch of post-processing to filter those out.
All Web Audio API, all client-side. No server in the audio path.